11(34)
uncertainty will lead to more cautious behavior (i.e., lower prices in the
case of procurement) but they cannot be said to be directly useful in a
practical procurement situation.
The assumption that the distribution of valuations is known to participants
is even more problematic. Quite apart from the epistemic problems
associated with the assumption as such, the actual behavior of participants
in itself proves that the assumption is wrong. If the distribution of values
and number of participants are known, it is possible (at least in principle,
assuming a basic knowledge of auction theory) to compute the price
level of the winning tender. It is then difficult to see why anyone should
go to the trouble of presenting a tender whose price level is far above
the winning level, given the cost of producing a tender. In one of the
procurements analyzed, the number of tenders was 27, and the expected
value for the tenderer with the highest price level was of the order of
10
-33
EUR. Clearly, such a tender disproves the assumption of known
distributions. In a more normal situation with 5 participants, no more than
3 tenders on average would in fact be submitted. In summary, the actual
behavior of producers in procurement situations is incompatible with the
standard model.
To these problems may be added that of estimating distributions even
ex post
. Laffont (1997) draws the conclusion that the outcome of such an
estimation depends strongly on the assumptions made concerning non-
observable distributions (see also Athey and Haile (2005) on the problems
of identifiability).
Klemperer (2002) makes the following general assessment of the academic
literature:
What really matters in auction design are the same issues that any industry
regulator would recognize as key concerns: discouraging collusive, entry-
deterring and predatory behavior. In short, good auction design is mostly
good elementary economics. By contrast most of the extensive auction
literature […] is of second-order importance for practical auction design.
The literature largely focuses on a fixed number of bidders who bid non-
cooperatively, and it emphasizes issues such as the effects of risk-aversion,
correlation of information, budget-constraints, complementarities, etc.
Auction theorists have made important progress on these topics which other
economic theory has benefited from, and auction theory has also been
fruitfully applied in political economy, finance, law and economics, labor
economics, industrial organization, etc. often in contexts not usually thought
of as auctions […]. But most of this literature is of much less use for actually
designing auctions.
Alternative Description
If the standard strategic model seems to be unrealistic in its assumptions
about the information available to tenderers and about their competence,
the most natural alternative is to maintain traditional assumptions about
rationality and profit maximization, while keeping in mind that decisions
on participation, pricing, and other parameters are made under genuine
uncertainty. It cannot be assumed that the number of tenderers or the
distribution of their valuations is known
ex ante
. Although tenderers are
aware of the existence of other tenderers, they are unable to adapt their
behavior, given that they know neither the identity nor the preferences of
